What stops it from spamming comments or closing real issues?

It stores maintainer overrides as ground truth for an accuracy report, applies labels and a single updated comment, and can notify Slack, Discord, or Telegram for critical or spam findings.

Will it auto-close anything it thinks is spam?

By default only high-confidence spam with allowAutoClose enabled is closed with a polite comment. Lower-confidence spam gets spam plus needs-triage and is never auto-closed.

Can I try triage logic without registering webhooks?

Yes. npm run triage with repo and number prints the decision and confidence, and adding apply writes labels and comments. Useful for iterating without webhooks.

How does duplicate detection avoid calling Jev for every issue?

Duplicate search does one list call with a local lexical prefilter to top candidates, then one noul Jev question per candidate for a 0..1 score, gated by maxCandidates and a comment threshold.

What does registering the GitHub App involve?

Node 20, MIT, Probot. Open the dev server, which redirects to /probot and offers Register a GitHub App; that flow creates the app, opens a smee.io channel and writes APP_ID, WEBHOOK_SECRET, PRIVATE_KEY and WEBHOOK_PROXY_URL into .env for you. Restart afterwards because Probot only reads .env at startup. Permissions needed are Issues read and write, Pull requests read and write, Contents read and Metadata read.

Which configuration knobs are actually available?

Every key is optional in .github/jev-triage.yml and validated with zod, with invalid values falling back to defaults: enabled, dryRun, allowAutoClose, triage.issues, triage.pullRequests, triage.onPullRequestPush, triage.includeDrafts, thresholds.actMin at 0.6, thresholds.duplicateComment at 0.85, duplicates.maxCandidates at 6, and ignore.labels defaulting to skip-triage. Labels named under labels: are created automatically the first time the bot applies them.

Which model does it call, and can I pin the version?

Pin it if your thresholds are sensitive, because jev-latest shifts behaviour between releases. Precedence runs most specific first: model in the repository's jev-triage.yml, then the JEV_MODEL environment variable as a deployment-wide default, then jev-latest. Every question and criterion is written in English on purpose, because Jev's primary training language is English and other languages currently score lower accuracy.

CommunityJev Tools & CLIs

Jev Triage — ashafizullah

Probot GitHub app that triages issues and pull requests with two Jev fan-outs for category, impact, completeness, spam, and duplicates, then labels and comments once.

See highlights

Open github.com

Stars: 1
Language: TypeScript
License: MIT
Published by: Community
Author: ashafizullah
Last verified: Oct 2026

ashafizullah/jev-triage

User repository on GitHub

View on GitHub

Automated issue & PR triage for open-source maintainers, powered by Jev (TypeSafe AI).

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

This is one of several Jev listings for the same project. The canonical entry is Jev Triage — boldbug1.

About this project

Automated issue & PR triage for open-source maintainers, powered by Jev (TypeSafe AI). An incoming issue or pull request is classified, scored, checked for missing information, compared against existing items for duplicates, and then labelled — with every uncertain decision routed to a human instead of guessed at. Classifies the item: bug, featurerequest, question, documentation, spam, other. Scores impact on a four-level rubric and derives a severity.

Quoted from the project's README. Read the full README.

Tagscommunity github

Works forJev for Classification

Used inJev for Software & DevTools

Languages: TypeScript

Highlights

Closed-set category, four-level impact, completeness, spam, and needs-human-review in one parallel fan-out.
Lexical prefilter plus per-candidate Jev noul verification for duplicate scoring.
Decision engine with confidence gates, composite spam risk, SQLite audit, and one updatable comment.
Zod-validated repo config, auto-created labels, dry-run mode, and optional chat notifications.

Watch out

Uncertain or risky calls get a needs-triage label for a human instead of being guessed; English questions only for best accuracy. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

Jevest

GitHub Action that reviews pull requests with a read-only finder agent plus a refuting verifier, using Jev only to triage, route, label, and gate merging.